Sorting of items on a moving conveyor belt. Part 2: performance evaluation and optimization of pick-and-place operations

Abstract

The fundamental issues to be addressed in automatic sorting systems are sensing, i.e., detecting and classifying the objects to be sorted, and gripping, i.e., realizing the required separation in the most efficient way. While the sensing issue is considered in Part 1 of this work, the topic of this paper is the evaluation and optimization of the gripping performance. For a sorting system to have a “good” behavior, it is required to pick up “the more it is possible”, in terms of percentage of gripped objects with respect to the total (gripping rate), and in terms of gripped objects per time unit (throughput). In the sorting of recyclable packaging (considered here as in Part 1 of this work), the flow of incoming items cannot be controlled and only its stochastic description may be known, so that the optimization of the above indices is a crucial and not trivial issue. Here, the problem is addressed at two different levels: first, different robot kinematics are characterized in terms of the parameters that affect the system performance, proving that the choice of a suitable kinematic structure is the first available tool to get a satisfactory system behavior. Then, for the class of redundant (or flexible) kinematic devices, the system performance is shown to depend on the strategy that governs the gripping order for the items on the belt. For this nonstandard scheduling problem, two innovative solutions have been devised, based on suitable modifications of standard queuing strategies. The proposed algorithms have the same complexity of the corresponding standard ones, but improved performance in the considered case, as confirmed by the reported simulation results.